This repository presents our work titled "Does Current Deepfake Audio Detection Model Effectively Detect ALM-based Deepfake Audio?," which was available on arxiv at "https://arxiv.org/abs/2408.10853".
The A03 VALL-E and A04 VALL-E X are same with the ALM subset (A1, A2) of Codecfake, which can be download at https://zenodo.org/records/11169781.
Pre-trained model can be found at "https://github.com/xieyuankun/Codecfake/tree/main/pretrained_model".
For the methods used to generate logits and calculate EER, please refer to logits generation method and EER calculation method.